Screening for Type 2 Diabetes
Report of a World Health Organization and
International Diabetes Federation meeting
World Health Organization
Department of Noncommunicable Disease Management
Screening for Type 2 Diabetes
Report of a World Health Organization and
International Diabetes Federation meeting
World Health Organization
Department of Noncommunicable Disease Management
2.1 Diabetes and its consequences..........................................................................1
2.2 Screening for type 2 diabetes – why WHO and IDF convened this meeting...2
2.3 Effects of screening on individuals, health systems and society......................3
2.4 Screening and prevention - the links ...............................................................4
3 Aims of the meeting .....................................................................................................5
4 Terminology – what is screening ..........................................................................5
5 Evaluating screening tests and programmes.................................................................6
5.1 General issues..................................................................................................6
5.2 Issues specific to diabetes................................................................................8
5.2.1 Range of available tests......................................................................8
5.2.2 Evaluating screening procedures........................................................8
5.2.3 Performance indicators.......................................................................9
5.2.4 Performance of screening tests for type 2 diabetes............................9
184.108.40.206 Urine glucose.......................................................................10
220.127.116.11 Blood glucose......................................................................11
18.104.22.168 Glycated haemoglobin.........................................................13
22.214.171.124 Combinations of tests..........................................................13
126.96.36.199 Public response to screening for type 2 diabetes.................14
188.8.131.52 Frequency of testing............................................................15
5.2.5 Assessing the risk of future development of type 2 diabetes...........15
6 The current evidence base..........................................................................................16
6.1 Evidence relating to the efficacy of early detection......................................16
6.2 Evidence relating to economic aspects of early detection.............................18
6.3 Evidence relating to the psycho-social effects of early detection.................20
7 Formulating policies about screening for type 2 diabetes..........................................21
7.1 The aims and objectives of a screening policy.............................................21
7.2 Epidemiological considerations....................................................................21
7.3 Considerations of health system capacity.....................................................21
7.4 Economic considerations..............................................................................22
7.5 The choice of a test or tests...........................................................................22
7.6 Competing priorities ....................................................................................23
7.7 Ethical and political considerations..............................................................23
8 Widening the evidence base .....................................................................................23
8.1 The need for evidence from randomized controlled trials............................23
8.2 The need for observational studies...............................................................24
8.3 The need for economic evidence..................................................................25
8.4 The use of modelling studies........................................................................25
8.5 The need for evidence on the psycho-social effects of early detection........25
9 Implementing policies about screening for type 2 diabetes......................................27
10 Conclusions and recommendations...........................................................................29
10.1 Conclusions................................................................................................ 29
10.2 Recommendations...................................................................................... 30
Annex 1 List of participants of the WHO/IDF meeting.............................................32
Annex 2 Acknowledgements......................................................................................33
Tables and figures..................................................................................................................35
Over the past decade it has been obvious that the prevalence of type 2 diabetes
is increasing rapidly. Unless appropriate action is taken, it is predicted that
there will be at least 350 million people in the world with type 2 diabetes by
the year 2030. This is double the current number. Equally alarming and less
well known is the fact that, of these people, only around one half are known to
have the condition. This has been shown repeatedly in epidemiological
surveys. An added concern is that half of those who do present with type 2
diabetes clinically already have signs of the complications of the disorder.
It has not yet been proven that earlier detection will improve the outcome of
people with type 2 diabetes, but it seems logical to suggest that it may help.
The implication of this is that people need to be screened for diabetes on a
regular basis. There is still uncertainty whether this should be done on a
population-wide basis or just for those people who can be shown to have a
high risk. It is also uncertain at what age the screening programmes should be
introduced, if at all.
This report focuses solely on screening for type 2 diabetes in non-pregnant
adults. It does not consider screening for type 1 diabetes, screening for type 2
diabetes in children, nor screening for gestational diabetes. This is not to
imply that these topics are unimportant. On the contrary, they are each
important enough to require detailed consideration in their own right.
It is clear to both the World Health Organization (WHO) and the International
Diabetes Federation (IDF) that guidance is needed for both our member
countries and member associations. Because of this the WHO and the IDF
have come together to produce this document, which, though it poses as many
questions as it answers, is a clear and logical start to a very serious debate.
We hope that the report will provide guidance and provoke discussion and
new studies and in the long term will be of benefit to the many people in the
world with and at risk of type 2 diabetes.
Dr Derek Yach Professor Sir George Alberti
Executive Director President
Noncommunicable Diseases International Diabetes Federation
and Mental Health Cluster
World Health Organization
2.1 Diabetes and its consequences
Diabetes mellitus is a metabolic disorder characterized by chronic
hyperglycaemia with disturbances of carbohydrate, fat and protein
metabolism resulting from defects in insulin secretion, insulin action,
or both1 . The current diagnostic criteria are shown in Table 1 In
summary, diabetes is diagnosed if the (venous) fasting plasma glucose
(FPG) value is >= 7.0 mmol l-1 (126 mg dl-1), or if the casual plasma
glucose value is >= 11.1 mmol l-1 (200 mg dl-1), or if the plasma
glucose value 2 hours after a 75g oral load of glucose >= 11.1 mmol l-1
(200 mg dl-1). In asymptomatic subjects, performing the test on one
occasion is not enough to establish the diagnosis (i.e. basis to treat
diabetes). This must be confirmed by carrying out at least one further
test on a subsequent day.
Impaired glucose tolerance (IGT) and impaired fasting glycaemia
(IFG) are risk categories for the future development of diabetes and
cardiovascular disease (CVD). An individual falling into the IFG
category on the fasting result may also have IGT on the 2-h value or,
indeed, diabetes. If an individual falls into two different categories, the
more severe one applies.
The classification of diabetes is based on aetiological types1 . Type 1
indicates the processes of beta-cell destruction that may ultimately lead
to diabetes in which insulin is required for survival. Type 2 diabetes is
characterized by disorders of insulin action and /or insulin secretion.
The third category, "other specific types of diabetes," includes diabetes
caused by a specific and identified underlying defect, such as genetic
defects or diseases of the exocrine pancreas.
The latest WHO Global Burden of Disease estimates the worldwide
burden of diabetes in adults to be around 173 million in the year 2002
. Around two thirds of these live in developing countries. Diabetes is
no longer a condition of developed, ‘industrialised’ or ‘Western’
countries. Global estimates of the burden of IFG and IGT are not
available, but the number of people with IGT is likely to be even
greater than the number with diabetes4,3. IGT and IFG are now
sometimes referred to as ‘pre-diabetes’ (a term not unanimously
supported by those attending this meeting since diabetes will not
necessarily develop in those with IGT or IFG).
The diabetes epidemic is accelerating in the developing world, with an
increasing proportion of affected people in younger age groups. Recent
reports describe type 2 diabetes being diagnosed in children and
adolescents5,6,7 . This is likely to increase further the burden of chronic
diabetic complications worldwide.
Most of the consequences of diabetes result from its macrovascular and
microvascular complications. (Some describe a third category –
‘neuropathic’, whereas others classify the diabetic neuropathies as
microvascular complications.) The age-adjusted mortality, mostly due
to coronary heart disease (CHD) in many but not all populations, is 2-4
times higher than in the non-diabetic population8 , and people with
diabetes have a 2-fold increased risk of stroke9 . Diabetes is the leading
cause of end stage renal failure in many populations in both developed
and developing countries10 . Lower extremity amputations are at least
10 times more common in people with diabetes than in non-diabetic
individuals in developed countries11 , and more than half of all non-
traumatic lower limb amputations are due to diabetes. In developed
countries, diabetes is one of the leading causes of visual impairment
and blindness12,13 .
People with diabetes require at least 2-3 times the health care resources
of people who do not have diabetes, and diabetes care accounts for up
to 15% of national healthcare budgets14,15 .
2.2 Screening for type 2 diabetes – why WHO and IDF convened this
The main reasons for the current interest in screening for type 2
diabetes and the reasons why WHO and IDF convened this meeting
• that there is a long, latent, asymptomatic period in which the
condition can be detected16,17 ;
• a substantial proportion of people with type 2 diabetes are
undiagnosed (Table 2);
• a substantial proportion of newly referred cases of type 2 diabetes
already have evidence of the micro-vascular complications of
• the rising prevalence19 of type 2 diabetes world-wide;
• the seriousness of the immediate effects and long-term
complications of type 2 diabetes;
• evidence supporting the efficacy of intensive blood glucose
control20,21 blood pressure control22 and blood lipid control23,24 , 25,26
in type 2 diabetes and
• accumulating evidence that treatment of hypertension,
dyslipidaemia (for example lowering LDL cholesterol23,24 ) can
prevent cardiovascular disease in people with type 2 diabetes.
• increasing pressure from professional organisations, lay groups and
from some of the member associations of IDF to institute screening
for type 2 diabetes if only to further highlight the increasing
prevalence and public health importance of the condition.
requests from national and regional health authorities and individual health care
professionals for guidance as to what should be their policies for screening for type 2
2.3 Effects of screening on individuals, health systems and society
Policies and practices for screening for type 2 diabetes have profound
implications for individuals, health systems and society as a whole.
Implications for individuals include:
• the time and other resources necessary to undergo the screening
test (or tests) and any subsequent diagnostic test (or tests);
• the psychological and social effects of the results whether the
screening test proves ‘positive’ or ‘negative’ and whether or not the
diagnosis of type 2 diabetes is subsequently made and
• the adverse effects and costs of earlier treatment of type 2 diabetes
or of any preventive measures instituted as a result of the individual
being found to have diabetes. These may include occupational
discrimination and/or increased costs or difficulty in obtaining
The effects on the health system and society as a whole are:
• the costs and other implications (especially in primary care and
support services such as clinical biochemistry) of carrying out the
screening test (or tests) and the necessary confirmatory test (or
• the additional costs of the earlier treatment of those found to have
diabetes or to be at high risk of developing diabetes or
cardiovascular disease in the future and
• the implications of false negative and false positive results which
are inevitable given that any initial test will be a screening test and
not a full diagnostic test (except in the case of an OGTT with
markedly abnormal values).
• any loss of production as a result of the earlier diagnosis of the
condition (from absence from work or reduced job opportunities,
The potential benefits of early detection of type 2 diabetes are:
• enhanced length and/or quality of life which might result from a
reduction in the severity and frequency of the immediate effects of
diabetes or the prevention or delay of its long-term complications.
• Any saving or redistribution of health care resources which might
be possible as a result of reduced levels of care required for
diabetes complications (reduced hospital admissions and lengths of
2.4 Screening and prevention - the links
Any programme aimed at the early identification of type 2 diabetes
through screening will also identify individuals with IGT and/or IFG.
Thus any policy, whether related to public health or day-to-day clinical
practice must specify what should be done when these conditions are
The prognostic significance of IGT and, to an extent IFG, is being
clarified27 . Also, evidence concerning the effect of interventions in
IGT is now available. In particular, interventions aimed at weight
reduction and increased physical activity and the use of some
pharmacological agents have been shown to be effective in reducing or
delaying the transition to diabetes in those with IGT.
In general, lifestyle interventions appear to be more effective than
medications 28 and the most important, recent trials published in peer
reviewed journals are summarised in Table 3.
The Diabetes Prevention Programme (DPP) Research Group evaluated
the cost-effectiveness of the interventions used in their trial29 . Life-
style intervention and metformin were both judged to be cost-effective.
The lowest value for cost per QALY gained (from the health system
perspective) was USD 8,100. This was for the comparison of lifestyle
changes with placebo, with lifestyle advice given as it might be in
routine clinical practice (i.e. less intensively) – to groups of 10 patients
- and with the optimistic assumption that there would be no reduction
in clinical effectiveness. The highest cost per QALY was USD 99,600.
This related to the comparison of metformin with placebo, as
implemented in the DPP trial (i.e. with individual clinical care). The
equivalent costs per QALY gained from the societal perspective were
USD 23,800 and USD 99,200.
Within the context of the US these were judged to be cost-effective. In
health care systems with lower staff and/or medication costs these
costs per QALY would be lower and the interventions, all other things
being equal, would be more cost-effective.
Given this new, encouraging information on the prevention of or delay
in the transition from IGT to diabetes, there is at least potential benefit
from the detection of this condition through screening. Whether
similar benefits will follow from the early detection of diabetes is
3 Aims of the meeting
The aims of this WHO/IDF meeting were:
• To review the scientific evidence for the usefulness of screening for early
detection of type 2 diabetes.
• To make recommendations relevant to health care policy, action and future
• To explain these recommendations in a joint WHO/IDF Report.
4 Terminology – what is screening?
The Group’s working definition of the term screening is based on that used in
the WHO "Principles of Screening" document 30 (September 2001 draft):
“Screening is the process of identifying those individuals who are at
sufficiently high risk of a specific disorder to warrant further investigation or
The definition goes on to say:
“It [screening] is systematically offered to a population of people who have
not sought medical attention on account of symptoms of the disease for which
screening is being offered and is normally initiated by medical authorities and
not by a patient's request for help on account of a specific complaint. The
purpose of screening is to benefit the individuals being screened.”
The term diagnosis refers to confirmation of diabetes in people who have
symptoms, or who have had a positive screening test. In diabetes, the
screening test may be the diagnostic test (e.g. a fasting plasma glucose => 7.0
mmol l-1 in someone who has symptoms) or the first part of the diagnostic test
if a second test (usually the OGTT) is used to confirm the diagnosis in
There are several potential approaches to screening for diabetes:
• Screening the entire population (never actually suggested since all
proposals have been, in some way, selective).
• Selective or targeted screening performed in a subgroup of subjects who
have already been identified as being at relatively high risk in relation to
age, body weight, ethnic origin etc.
• Opportunistic screening carried out at a time when people are seen, by
health care professionals, for a reason other than the disorder in question.
‘Selective or targeted screening’ and ‘opportunistic screening’ are not
mutually exclusive since screening may be limited to those at highest risk. In
opportunistic screening, the decision to initiate the health care encounter is
made by the individual, albeit for reasons not related to the condition for
which screening is offered. This needs to be distinguished from screening
programmes in which the invitation to come forward and be screened is part of
There is also ‘haphazard’ screening, characterised by a lack of a coherent
screening policy. In such cases individuals may be invited to be screened
irrespective of their risk (people in a supermarket, for example) or there may
be no adequate explanation of the reasons for screening or no formal system of
support for those taking part, whatever the outcome of their test.
5 Evaluating screening tests and programmes
5.1 General issues
The sensitivity of a screening test is the proportion of people with the
disorder who test positive on the screening test. (A highly sensitive
screening test is unlikely to miss a subject with diabetes.)
The specificity of a screening test is the proportion of people who do
not have the disorder who test negative on the screening test. (A highly
specific test is unlikely to misclassify someone who does not have
diabetes as having diabetes.)
Although it is desirable to have a test that is both highly sensitive and
highly specific, this is usually not possible. In choosing a cut-off point
a trade-off needs to be made between sensitivity and specificity, since
increasing one reduces the other. The receiver operator characteristic
(ROC) curve expresses this relationship. The true positive rate
(sensitivity) is plotted on the y axis against the false positive rate (1-
specificity) over a range of cut-off values. Tests that discriminate well
crowd toward the upper left corner of the ROC curve (Figure 1). In
ideal cases, as sensitivity increases, there is little decrease in
specificity, until very high levels of sensitivity are reached31 .
What should happen, in practice, is that ROC curves should be used in
conjunction with pre-specified performance indicators (such as the
proportion of cases that should be identified, what proportion of re-
tests are acceptable). Some measure of ‘trade-off’ between
performance indicators is likely to be necessary.
Validity is the extent to which the test reflects the true status of the
Reliability is the degree to which the results obtained by any given
procedure can be replicated.
Reproducibility refers to obtaining similar or identical results on
repeated measurements on the same subject.
Screening tests must be shown to be valid, reliable and reproducible in
the population in which screening is to take place. Uniform procedures
and methods, standardized techniques, properly functioning equipment,
and quality assurance are all necessary to ensure reliability and
Predictive value relates to the probability that a person has or does not
have the disorder given the result of the test. Thus:
Positive predictive value is the probability of the disorder in a
person with a positive test result and negative predictive value
is the probability of a person not having the disorder when the
test result is negative.
The predictive value of a test is determined not only by the sensitivity
and the specificity of the test, but also by the prevalence of the disorder
in the population being screened. Thus, a highly sensitive and specific
test will have a high positive predictive value in a population with a
high prevalence of the disorder. This is part of the rationale for
promoting selective or targeted screening. When the prevalence is low,
as may be the case when the entire population (or the entire adult
population) is screened, then the positive predictive value of the same
test will be considerably lower. In this case, a high specificity drives a
high positive predictive value. To avoid false positives (throughout the
range of prevalence) it may be necessary to increase specificity at the
expense of sensitivity.
Screening tests may be used in parallel (i.e. a person is deemed to be
likely to have a disorder if they test positive to either test). In this case
the sensitivity and the negative predictive value are generally increased
and the specificity and positive predicted values decreased.
On the other hand, screening tests may be used in series (i.e. a person
needs to be positive to both tests in order to be deemed likely to have
the disorder). In this case the specificity and positive predicted value
are generally increased and the sensitivity and negative predicted value
decreased. Tests in series have been advocated in type 2 diabetes (this
is further discussed below) when, for example, a questionnaire may
precede a fasting blood sample or OGTT and be used to exclude some
individuals deemed to be at low risk of having the disorder.
5.2 Issues specific to diabetes
5.2.1 Range of available tests
Screening tests for type 2 diabetes include risk assessment
questionnaires, biochemical tests and combinations of the
two. The biochemical tests currently available are blood
glucose or urine glucose measurements, blood HbA1c or blood
fructosamine measurements. Each screening test needs a
designated and pre-determined threshold or ‘cutpoint’ that
defines high risk. Screening tests are usually followed by
diagnostic tests (fasting plasma glucose (FPG) and/or an oral
glucose tolerance test (OGTT) using standard criteria) in order
to make the diagnosis.
5.2.2 Evaluating screening procedures
Meaningful evaluation and comparison of the performance of
screening tests and procedures for diabetes should be carried
out against specified criteria and should take into account the
following basic principles:
• People with known diabetes should not be included in
the prevalence data used to calculate PPV
• Selection of cut-off points:
o should ideally be determined using ROC curve
analysis because this considers performance
over the whole range of cutpoints
o alternatively these can be determined by using a
common specificity or sensitivity
o should take into account the aims of the
screening programme, available resources to
meet the workload which will be generated by
the proportion of the population which will
require further testing, and the importance
placed on avoiding false positive and false
• A valid assessment of screening tests requires the whole
screened population (or a sample of them) to have
diagnostic testing, not just those who screen positive
• Performance should be validated on a population
different to that from which the screening procedure
• A distinction should be made between an
epidemiological and a clinical diagnosis of diabetes. An
epidemiological diagnosis can be based on a single
OGTT or FPG whereas a clinical diagnosis, in the
absence of symptoms, requires confirmation by a repeat
• The precisely specified objectives of the programme.
5.2.3 Performance indicators
A standard set of performance indicators should be used to
evaluate a screening procedure or test and include: statistical
performance (sensitivity, specificity, PPV, ROC - area under
the curve) and the percentage of the population identified which
requires further or definitive testing. Additional indicators
include information on the cardiovascular disease risk profiles
of identified individuals and measures of the economic
performance of screening tests and population measures such as
the acceptability of the screening programme to those invited to
attend, the extent to which any lack of acceptability reduces
uptake, the psychosocial impact of each screening outcome –
positive and negative, ‘true’ and ‘false’ and the ability of those
found to be at risk of future development of diabetes to modify
5.2.4 Performance of screening tests for type 2 diabetes
These have been recently extensively reviewed32-35. Some
caution is required in interpreting the statistical results reported
in these reviews and below because in many studies the
diagnosis of diabetes was made using diagnostic criteria which
predate the current WHO and American Diabetes Association
(ADA) criteria. Despite this, the data allow conclusions about
general performance of the various approaches to screening for
type 2 diabetes.
Several questionnaires have been developed to screen
for undiagnosed diabetes and have included a range of
questions covering both symptoms and recognised risk
factors. If a person presents as a result of any of the
symptoms of diabetes (such as thirst, polyuria etc.) and
is confirmed to have the condition then this process is
diagnosis and not screening. However, it is conceivable
that people identified as having diabetes by means of a
screening test or programme may, retrospectively,
recognise the presence of symptoms which were not
acted upon at the time. However, since the main
purpose of screening is to detect asymptomatic people
with undiagnosed diabetes, questionnaires which are
based on the symptoms of diabetes are not considered
The original ADA “Take the test: know the score”
questionnaire36 included both symptoms and historical
risk factors. A modified version of this questionnaire
has been evaluated by Herman et al37 based on data
from the Second National Health and Nutritional
Examination Survey and had a sensitivity of 83%,
specificity of 65% and PPV of 11%. This questionnaire
was subsequently tested in a community screening
program in Onondaga County New York and showed a
sensitivity of 80%, specificity of 35% and a PPV of
12% 38 ,39 .
Griffin et al developed a risk score based on risk factors
routinely collected in clinical practice40 and evaluated
this in a hypothetical notional population with the same
age-sex structure as England and Wales. No cut off for
the risk score was prescribed but rather criteria for
deciding a suitable cut point were proposed. An
example gave a sensitivity of 77%, specificity of 72%
and PPV of 11%.
184.108.40.206 Urine glucose
The usefulness of urinary glucose as a screening test for
undiagnosed diabetes is limited because of the low
sensitivity which ranged from 21% to 64% with
specificity > 98% in studies which included performing
an OGTT in the entire study population or a random
sample of negative screenees32 .
Examples of such studies include Davies et al41 who
used a self-test for postprandial glycosuria and reported
a sensitivity of a positive urine test of 43% and
specificity 98%. Hanson et al42 studied Pima Indians
with non-fasting urine glucose and non-fasting OGTT
and reported a sensitivity of 64% and specificity of 99%
for a positive urine test for diabetes diagnosed on the 2-
hour non-fasting post glucose load plasma glucose
Friderichsen and Maunsbach43 screened 2,242 people
with a self-test for postprandial glycosuria and tested all
people with a positive result and a random sample of
106 negative screenees with an OGTT and reported a
sensitivity of 21% and specificity of 99%.
Despite its low sensitivity, urine glucose testing may
have a place in low resource settings where no other
procedure is possible. This is particularly so, of course,
when the prevalence of undiagnosed diabetes is likely to
220.127.116.11 Blood glucose
Many studies of this question have used the blood
glucose measurement which was part of the diagnostic
test. In addition many studies have included people with
diagnosed diabetes in the statistical analysis of test
performance. Only studies which excluded people with
diagnosed diabetes are considered below.
Venous fasting plasma glucose32 has a sensitivity
between 40% and 65% with a specificity > 90% for
FPG values ranging from 6.1 – 7.8 mmol l-1. Since the
introduction of the new WHO and ADA diagnostic
criteria for diabetes attention has focussed on
comparisons of the cutoff point between normal and
abnormal - FPG of 6.1 mmol -1 as recommended by
ADA and 5.5mmol l-1, being the WHO cut point below
which the diagnosis of diabetes is unlikely. Although a
number of studies have supported the lower FPG value,
there is no universal agreement on this point and
ultimately the choice of cutpoint must be determined by
the purpose of the screening programme and the
resources required and available to perform further
testing on the proportion of the population which would
be identified by the choice of cutpoint33 .
Examples of studies which reported optimal sensitivity
and specificity at the lower cut-point include Costa et
al44 who reported that this was achieved at an FPG ≥ 5.4
mmol l-1, Larsson et al45 at an FPG of 5.3 mmol l-1 and
Cockram et al46 at an FPG of 5.6 mmol l . Modan and
Harris47 compared various FPG levels in people in the
USA and Israel and concluded that no FPG level
provided a satisfactory cutoff point to use in screening
for undiagnosed diabetes. However an FPG of ≥ 5.55
mmol l-1 performed better than other levels with a
sensitivity of 83% and 95% respectively in the USA and
Israel with corresponding specificities of 76% and 47%,
and PPVs of 17.2% and 11.8%.
A number of studies in different populations have
reported on the performance of an FPG of 6.1mmol l .
They report sensitivities ranging from 58%-87%
(median – 81%) and specificities ranging from 75%-
98% (median 92%)33 .
Fasting capillary blood glucose has also been used for
screening. Bortheiry et al48 reported that the best
equilibrium between sensitivity and specificity for the
diagnosis of diabetes was achieved at a cutoff of 5.6
mmol l-1 for fasting capillary blood glucose in their
study of 4,019 Brazilian people undergoing an OGTT.
Studies on the usefulness of random blood glucose
(RBG) as a screening test have mostly used random
capillary blood glucose (RCBG) measured with a blood
glucose meter. Two interrelated issues arise (1) the
usefulness of RBG/RCBG in screening and (2) the
accuracy of blood glucose meters for use in screening
There are few well designed studies which have
properly addressed these issues with the main
methodological problem being the failure to determine
the overall prevalence of undiagnosed Type 2 diabetes
in the cohort being studied by performing an OGTT in
everyone or in a sample.
Three studies have examined RCBG (measured by
reflectance meter) as a screening test for diabetes and
performed an OGTT in the whole population
irrespective of the RCBG result. Qiao et al49 reported a
sensitivity of 79% in men but only 40% in women while
specificity was 86% and 84% respectively for men and
women using for a value of 5.8 mmol l . Engelgau et
al reported that a value of 5.6 mmol l-1 achieved a
sensitivity ranging from 68%-74% and specificity
ranging from 66%-77% depending on age. The authors
concluded that different cutpoints are required to
account for the postprandial period and age. Rolka et
al51 reported that RCBGs of ≥ 7.8mmol l-1 and ≥
6.7mmol l-1 had sensitivities of 75% and 56% and
specificities of 88% and 96% respectively.
The other important consideration with blood glucose
meter measured readings is the accuracy of the result.
While these are sufficiently accurate for day to day
monitoring of diabetes control, their accuracy in
screening for undiagnosed diabetes in routine practice,
as opposed to the carefully controlled situations which
have applied in published studies, has been questioned
and attention drawn to the potential inherent inaccuracy
of the method52 . However these problems are not
necessarily insurmountable and the final decision about
their use will depend on resources and related practical
18.104.22.168 Glycated haemoglobin
The desire to replace the OGTT with a simpler test has
been a major factor behind the evaluation of glycated
haemoglobin as a screening test for undiagnosed type 2
diabetes. Peters et al53 performed a systematic review of
articles published between 1966 and 1994 in which
glycated haemoglobin (HbA1c) was measured
concurrently with performing an OGTT. When an
HbA1c plus 4SDs was used as a cutoff point, the
sensitivity was 36% and specificity 100% compared
with OGTT diagnosed diabetes using 1985 WHO
criteria. An HbA1c cutoff point of 6.1%, which had been
found to correspond most closely with a 2-hour plasma
glucose concentration of 11.1 mmol -1 in Pima Indians,
included 41% of non-diabetic subjects and 21% of
subjects with IGT l53 .
Davidson et al54 examined HbA1c levels from the
NHANES III study and from the Meta-Analysis
Research Group cohorts52 . Using the 1997 ADA criteria
for diagnosing diabetes, 60% of people in both datasets
having diabetes diagnosed on the basis of an FPG of
7.0-7.7 mmol l had normal HbA1c and one third had
results within 1% of the upper limit of normal.
Measurement of HbA1c is not yet standardised around
the world and has significant biological variation in
non-diabetic subjects55 . There is currently insufficient
evidence to enable a judgement to be made regard to its
performance as a screening test.
Fructosamine has been used less frequently than
glycated haemoglobin measurement, and has not
performed well because of sensitivities as low as 19%32 .
22.214.171.124 Combinations of tests
Screening tests, as mentioned above, may be combined
to improve performance. In relation to type 2 diabetes
this can be done using the tests serially (e.g. assessing
risk by questionnaire followed by blood glucose
measurement if a certain risk score is reached) or
simultaneously (e.g. measurement of blood glucose and
HbA1c at the same time). Combination testing is more
resource intensive, especially if applied sequentially.
Tests performed in parallel using FPG and HbA1c or
fructosamine have been reported to have a sensitivity
ranging from 40% to 83% and specificity of 83%-99%,
depending on the cut off values chosen32 . Combining
the modified ADA questionnaire and RCBG ≥ 6.7
mmol l-1 achieved a sensitivity of 58% and specificity of
An illustration of the effects of serial combination
testing is shown in Table 4 for a screening protocol
which initially assessed risk factors, performed FPG in
those at risk, then measured HbA1c in those with an
FPG between 5.5 and 6.9 mmol l-1, and then tested with
an OGTT those who had an HbA1c ≥ 5.3%. This
example illustrates that serial testing results in
decreasing sensitivity, increasing specificity and PPV
and reduces the number of people requiring definitive
Multivariate logistic regression modelling with
derivation of a probability value is another approach to
combining demographic, clinical and biochemical tests
in screening for undiagnosed diabetes. Tabaei and
Herman56 combined age, sex, BMI, postprandial time
and random capillary plasma glucose to calculate the
probability of undiagnosed diabetes and therefore the
need for an OGTT. The calculation can be performed on
a hand held programmable calculator and had a
sensitivity of 65%, specificity of 96% and PPV of 63%.
126.96.36.199 Public response to screening for type 2 diabetes
Not much is known about the public response to
diabetes screening programmes. This is important in
that the sensitivity of a screening programme, even if
based on a test or tests of optimum sensitivity, will be
severely reduced if uptake is low. Epidemiological
studies of diabetes prevalence in the general population
could serve as proxies to provide this information, but
only prevalence studies with a high response rate are
reported in the literature.
In the few available studies that do report the response
to a diabetes screening programme, it is in the range of
30% - 80%57-60 . In all these studies the population
invited for screening was defined by a specific age
range only. There are no reports on whether the
response of these populations would have been different
if a higher-risk population were invited, defined by
several known risk factors (age, obesity, family history
188.8.131.52 Frequency of testing
There are no compelling data on which to decide the
optimum frequency of screening for type 2 diabetes.
One possible source of information could be studies of
diabetes incidence or progression to diabetes from
normal glucose tolerance, IGT or IFG.
The annual rate of progression from IGT and IFG to
diabetes is 3% - 13% 61,62 , which might argue for annual
re-screening in people with IGT and/or IFG. However,
there are fewer studies of the incidence of diabetes in
normoglycaemic individuals. The available data
indicate that the annual progression from
normoglycaemia to diabetes is in the range of 0.6% -
1.2%, depending on the population and the age group
On the basis of the available data, the ADA have
recommended screening of middle-aged
normoglycaemic individuals at 3-yearly intervals63 ,
while the British Diabetic Association (now Diabetes
UK) recommended screening of 40-75 year-olds every
5 years if they have none of the recognised risk factors,
and every 3 years in the presence of risk factors64 .
5.2.5 Assessing the risk of future development of type 2 diabetes
In addition to the detection of undiagnosed type 2 diabetes there is
increasing interest in identifying people without diabetes who are at
increased risk of the future development of the condition. Any
screening strategy which aims to identify people with undiagnosed
diabetes which includes an OGTT will identify some people with IGT
and IFG and screening strategies which use an FPG will identify
people with IFG.
Stern et al used a multiple logistic regression model to predict 7.5 year
incidence of type 2 diabetes using readily available clinical
information65 . The model included age, sex, ethnicity (Mexican
American/non Hispanic white), FPG, systolic blood pressure, HDL
cholesterol, BMI and parent/sibling with diabetes. This model
performed similarly well by area under the ROC curve with and
without inclusion of the 2h OGTT blood glucose value.
The Finnish Risk Score is a recent example of a risk assessment tool
designed to identify people at risk of the future development of
diabetes without the need for laboratory tests66 . The risk score included
age, BMI, waist circumference, history of antihypertensive drug use or
high blood glucose, physical activity and daily consumption of fruit
and vegetables. A score of ≥ 9 had a sensitivity of 77% and specificity
of 66% and PPV of 7% in a large cohort followed for 10 years.
6 The current evidence base
6.1 Evidence relating to the efficacy of early detection
There are currently no definitive RCT data available on the efficacy of
early detection of diabetes through screening. A number of relevant
studies are in progress or have recently reported, albeit with weaker
designs than RCTs.
‘INTER-99’ is a study currently taking place in Copenhagen county,
Denmark. In this study, over 13,000 residents aged 30-59 were invited
to take part in a screening programme and were randomised to
intervention (90%) and control (10%) arms. Just over 50% (6,784)
accepted the invitation and were tested with a standard 75 g OGTT. In
the intervention group, people with newly diagnosed diabetes, IGT or
at high risk of CVD (the upper quintile of risk) were invited to
participate in a lifestyle modification programme concentrating on
reducing fat and energy consumption and increasing physical activity.
In addition, people received ‘usual care’ with respect to medication
through their local practitioners. In the control group no specific
action was taken and these people receive only usual care. The results
are likely to be available in 2006.
The (Anglo-Danish-Dutch) ADDITION study67 is screening over
200,000 individuals aged 40-69 years for diabetes in a step-wise
screening strategy. It first identifies high risk individuals by means of
a risk score and then confirms or otherwise a diagnosis of diabetes
based on glucose concentrations at fasting and/or 2h after a 75g
glucose load. As a result of this process, a predicted 3,000 individuals
will be randomised to standard treatment or target-driven intensive
pharmacological and non-pharmacological treatment. The specified
end-points are all-cause mortality and fatal or non-fatal macrovascular
events. The study evaluates the effectiveness of screening and
intensive treatment with regard to these primary endpoints, the
secondary endpoints of microvascular complications, the cost-
effectiveness of this approach and psychosocial factors such as stress
and anxiety related to screening and subsequent treatment.
The study by Schneider et al68 carried out in the former German
Democratic Republic provides an analysis of a mass-screening
programme based on urinary glucose levels, conducted in the former
East Germany in the 1960’s and 1970’s. It suggested that those found,
by screening, to have diabetes had an improved outcome compared
with those presenting spontaneously with diabetes. However, the
methods used in this study would not be acceptable given current
criteria for RCTs.
In a post-hoc analysis of UKPDS (UK Prospective Diabetes Study)
data the frequency of subsequent complications in relation to FPG on
entry to the trial has been carried out 69 . The rationale behind this
analysis is that FPG concentration on entry might serve as a surrogate
for the duration of diabetes prior to recruitment. No significant
difference in the frequency of end-points between ‘incidental’ and
‘non-incidental’ cases with FPG 10 mmol l-1 and above was seen.
However, a significantly lower rate of all major end-points was seen in
the group with initial FPG <7.8 mmol l-1 compared with the =>10
mmol l-1 group and significantly lower diabetes related death rates and
myocardial infarction rates when the <7.8 mmol l-1 group was
compared with the <7.8 – 10 mmol l-1 group.
Although the UKPDS was not intended to test any a priori hypothesis
related to the early detection of diabetes, these findings do suggest a
benefit of intervention either at lower levels of FPG or at earlier stages
of the natural history of diabetes and this may be consistent with a
benefit derived from early detection. This latter inference is crucially
dependent on whether or not the value of FPG is an indicator of prior
duration of disease or simply a marker of the severity of the disease.
One of the purposes of screening for diabetes may be to reduce the risk
of cardiovascular disease in people with hyperglycaemia.
People with IGT and diabetes are at greatly increased risk of
cardiovascular disease, including heart disease, stroke and peripheral
vascular disease70 . This risk appears to be associated with both
hyperglycaemia71 and to an increased frequency of other recognised
cardiovascular risk factors in people with diabetes72 . Cardiovascular
risk does vary according to the diagnostic criteria used, with glucose
tolerance apparently more strongly associated with risk than fasting
glucose levels 73 . Glycated haemoglobin levels also predict
cardiovascular risk in non-diabetic as well as diabetic individuals74 .
There is also now evidence that drugs that reduce cardiovascular risk
are highly effective in people with diabetes (for example
antihypertensives75 , statins76 ).
Diabetes screening may therefore be considered as one element of a
wider programme to identify individuals at increased cardiovascular
risk who could benefit from pharmacological treatments and lifestyle
change to reduce their cardiovascular risk. In practice this integration
often already occurs where cardiovascular disease is a significant
public health issue and policy for primary or secondary prevention of
cardiovascular disease has been developed. For example, WHO
guidelines recommend screening for diabetes in individuals with
hypertension77 and the National Service Framework for Coronary
Heart Disease in the United Kingdom recommends screening for
diabetes in individuals with documented cardiovascular disease78 .
The appropriate relationship between cardiovascular risk reduction
programmes and screening for diabetes will depend on local
circumstances. If there is already a cardiovascular risk reduction
programme, then the most cost-effective way of introducing screening
for diabetes is likely to be an integrated programme that identifies
individuals at high risk of both diabetes and cardiovascular disease and
offers appropriate pharmacological and lifestyle interventions for both.
This may influence the choice of population screened (targeting older
individuals with other cardiovascular risk factors such as hypertension,
for example). It may also influence the choice of screening test (if a
fasting sample is taken it could be tested for both lipids and glucose).
In populations where cardiovascular disease is not a major cause of
morbidity and screening is more appropriately directed at reducing
microvascular complications, screening may target younger individuals
who are at lower cardiovascular risk but do have a high lifetime risk of
microvascular complications. Similarly the choice of a glucose
tolerance test or fasting glucose measure to identify diabetes may
partly depend on whether the focus is on reducing cardiovascular or
6.2 Evidence relating to economic aspects of early detection
Descriptive data on costs suggest that the health care costs of the
screening itself are relatively low though there may be a substantial,
and as yet unquantified, opportunity cost both to the system and to the
individuals concerned. The costs of the subsequent treatment of
diabetes are likely to be much higher than the screening costs. As the
effectiveness of the management of diabetes becomes greater, the
relative benefits of early detection will become less.
Without direct evidence of the effectiveness of early intervention, there
can be no definitive statement of its cost-effectiveness. However,
modelling studies have been undertaken. For example, in the USA, the
CDC Diabetes Cost Effectiveness Study Group published the results of
an economic appraisal comparing the life time cost and benefit of
opportunistic screening for type 2 diabetes in adults with the then
current clinical practice79 . This group considered only direct medical
costs from a single payer’s perspective. The benefit was analysed by
using a computer simulated model incorporating the benefit of
preventing and reducing long term micro vascular complications. The
results demonstrated that opportunistic screening increases the lifetime
costs of treatment by $3,388 but results in a gain in life-years of only
0.02 years (1 week). To gain an additional life-year, the incremental
cost of screening over current clinical practice was estimated at
$236,449, and the cost per QALY gained was estimated at $56,649.
These same authors also compared the cost-benefit ratio of
opportunistic screening for diabetes with that of breast screening for
women aged 50 and above costing $34,000-$83,830 per life year
gained, cervical screening for women above 21 years of age costing
$50,000 per life year gained and hypertension screening for adult men
and women costing $48,000 and $87,000 per life year gained
respectively. Screening for type 2 diabetes thus compares
unfavourably with these other options.
In Taiwan, Chen et al80 used a computer simulation model to estimate
the cost per QALY and life years gained comparing two mass
screening programmes at two and five year intervals. The costs
compared included cost of screening and clinical care of diabetes and
its long term complications. The population disease progression model
used in the computer was derived from studies in the local Taiwan
population with, at that time, a prevalence of diabetes of 6%-12%.
Unlike the CDC simulation mentioned above, the authors included the
possible effect of reduction in macro vascular complications in their
This Taiwan simulation estimated a much lower cost of mass screening
programme per life year gained compared to the opportunistic
screening costs estimated by the CDC group. In Taiwan, the
incremental costs for biennial screening regime were estimated at
$26,750 per life-year gained, and $17,833 per QALY. The
corresponding figures for five-yearly screening regime were $10,531
per life-year gained and $17,113 per QALY. Although this comparison
is interesting, it would be misleading to draw firm conclusions from it
since the assumptions made were not identical in the two studies and
they relate to different populations. One particular difference between
the models is whether the benefit of reducing macrovascular
complications is or is not included.
In the absence of direct evidence on the effectiveness and cost-
effectiveness of screening for type 2 diabetes, simulations such as these
which are totally explicit in the assumptions made, can prove useful
guides as to the utility of screening in any given population.
6.3 Evidence relating to the psycho-social effects of early detection
This is an under-researched area in general and in particular in relation
to type 2 diabetes. Johnson et al81 report on levels of anxiety in
relation to screening for type 1 diabetes. Parents of ICA positive
children were reported to have the highest levels of anxiety followed
by spouses of positive individuals with at risk children having the
lowest levels of all.
Studies of conditions related to diabetes – hypertension and CVD are
also available. A systematic review of the psychological impact of
predicting individual risk of illness82 included 21 studies related to
CVD as well as two related to type 1 diabetes. Overall, the results
showed a range from increased anxiety, depression or psychological
distress through no effect to decreased levels. People undergoing
multi-channel chemistry screening83 generally show a short term
increase in anxiety which dissipates over time. Abnormal results are
associated with significant decrease in levels of perceived physical
health, general health, perceived health and pain. The people most
affected are those not offered any support.
More recently, Griffin et al84 have suggested that screening for
diabetes does not cause adverse psychological effects provided
appropriate explanations of the procedure are given and provided there
is appropriate follow-up. These conclusions have been strengthened
by recent studies from The Netherlands85 and from the USA86 . The
former interviewed 40 subjects involved in the Hoorn ‘stepwise’
population project which screened for diabetes. Twenty of these
people had been found to have previously undiagnosed diabetes and 20
were at increased risk of having diabetes but did not, at least at that
time, meet the diagnostic criteria for the condition. Although (or
perhaps because) the newly diagnosed subjects had little understanding
of the relevance of their diagnosis, only one was alarmed by it. Both
groups expressed positive views about the screening. In the US
study86 , screening for type 2 diabetes at the Durham Veterans Affairs
Medical Center identified 56 people with previously unknown diabetes
(out of 1,253 45-64 year olds screened) who had similar SF-36
Physical Component Scores and Mental Component Scores to those
who did not have diabetes, both at the time of screening and after one
Clearly these findings need to be confirmed with other subjects in other
cultures. As public awareness grows of the significance of a diagnosis
of diabetes and its possible long term consequences, the psychological
effects may be more marked.
7 Formulating policies about screening for type 2 diabetes
Figure 1 summarises the issues which need to be taken into account when
formulating a screening policy. Three epidemiological considerations have
been included, four of health system capacity and two economic
considerations. Each of these should be viewed as a spectrum from, at one
end, a clear indication that screening should be instituted to, at the other, clear
evidence that it should not. Any population at any one time will be at a given
point along each of these spectra. The policy decision as to whether or not to
institute screening will be a judgment which cannot, necessarily, be
extrapolated to other situations.
In reaching this judgement, public health authorities, clinicians, diabetes
associations and others should consider the following:
7.1 The aims and objectives of a screening policy
Aims should be clear and relevant to the context of screening
individuals at risk of having undiagnosed diabetes or at risk of
developing diabetes. These may relate to be the immediate effects of
diabetes (e.g. infections), to the prevention of microvascular
complications, to the prevention of CVD or to a combination of these.
Thus, of crucial importance in relation to framing aims and objectives
is knowledge about the most important consequences of diabetes in the
population being considered.
7.2 Epidemiological considerations
The most important epidemiological consideration is the prevalence of
undiagnosed type 2 diabetes; this is known for some countries and
regions as a result of field surveys of diabetes using OGTT; where it is
unknown, estimates can be made by extrapolation (ratio of previously
undiagnosed to known diabetes is likely to be around 1:1 or 2:1 (Table
2) but might be as extreme as 1:2 (e.g. Brazil) or 6:1 (Tonga). If
unknown, this can be determined by a relatively simple survey. If
screening of any kind is initiated, data on the numbers of unknown
cases identified need to be collected and analysed periodically.
7.3 Considerations of health system capacity
The main issues here are – the capacity of the system to carry out
screening, follow-up and diagnostic testing and its capacity to manage
effectively the newly detected cases of diabetes. The system must also
be able to support individuals when the results of the screening are
known, whether true positives, false positives or false negatives. The
identification of people with IGT and IFG is a by-product of screening
for type 2 diabetes and any screening policy needs to specify a clear
care pathway for such people Other issues are the capacity of the
system to assess individual risks by using routinely available data. The
first ‘screen’ might be made from such data87 thus eliminating one
attendance by the individual.
7.4 Economic considerations
The cost of a screening programme will vary depending on the costs of
materials, personnel etc. in any given setting. A large determinant of
cost is whether the activity is conducted as a de novo activity or builds
on existing health (e.g. primary care) infrastructure. Clearly, the
former is more costly. Also varying costs in getting the person to the
test. Cost of subsequent care will also vary widely (e.g. $10,000/yr in
USA vs. $100/yr in India. It is the second of these (cost of subsequent
care) which is the larger component of cost. May be difficult to build a
holistic economic argument since direct costs and indirect costs fall on
different public or private sectors and the funding of screening and
subsequent treatment may come from different budgets lines.
Costs of screening may be reduced if screening uses routinely available
information (for example using routine clinical information systems to
identify people at high risk of diabetes) or by linking to other screening
programmes (for example screening for glucose and lipids on the same
fasting blood sample as part of a cardiovascular screening programme).
Cost-effectiveness of treating screened individuals may also be
increased by screening populations at particularly high risk of
preventable adverse outcomes, for example populations with a high
risk of cardiovascular disease. However, even the screening of high
risk populations, such as those already known to have hypertension and
dyslipidaemia88 , can be costly. O’Connor et al88 identified 1,548 such
patients being cared for by a large medical group in Minnesota. After
exclusion of those who had already been screened in the past year,
those already known to have diabetes, those lost to follow-up (died or
left the scheme etc.), one newly diagnosed case was discovered for
every 40 high risk patients screened. This low yield resulted in a cost
of US$ 4,064 per case identified.
7.5 The choice of a test or tests
This is a judgement dependent upon the characteristics of a test
(sensitivity, specificity etc.) the cost of a test in any given context and
the capacity of the system to apply the test. Costs are largely driven by
the rate of detection of positives (true and false) and the need to follow
these up and carry out diagnostic testing.
7.6 Competing priorities
There will be competing priorities within diabetes – increased care for
people with known diabetes v identifying new cases; also in the wider
health care context - other health priorities such as communicable
diseases; these need to be reassessed on an ongoing basis.
7.7 Ethical and political considerations
Valid argument for considering it the right of any individual to have
diabetes diagnosed; this right needs to be weighed against any harm,
anxiety etc. that may occur as a result of earlier diagnosis; also the
harm done to false positive and false negative individuals; there is an
opportunity cost if screening is carried out because the resources
devoted to screening cannot be used for other purposes. No screening
programme can be instituted without a political will to do so. The
political will to institute screening may run counter to the supporting
Different expectations and ethical imperative depending on how the
person comes to the test: disease – patient comes to seek advice
screening – health professional imposes something on patient also
medico-legal ramifications of not screening.
8 Widening the evidence base
8.1 The need for evidence from randomized controlled trials
The benefits of early detection of type 2 diabetes through screening are
not clearly established. The few available studies suffer from several
types of bias that may lead to spurious conclusions regarding the
benefits of screening .
The main sources of bias are as follows:
Lead-time : the interval between the time of detection by screening and
the time diabetes would have been diagnosed in the absence of
screening. Thus lead-time bias prolongs the apparent duration of
survival and/or complication-free period simply by advancing the
Selection bias: People who enter screening programmes are volunteers
who are almost always more health conscious than the rest of the
population. Thus they are more likely to have a better disease outcome
even without screening.
Length-time bias : This relates to the fact that individuals with rapid
metabolic deterioration will tend to develop symptoms that prompt
them to contact health services. Thus only people with slowly
progressing and milder disease remain to be identified by screening.
These people are likely to have a better clinical outcome than rapidly
progressing cases, regardless of the treatment.
Over-diagnosis bias occurs when enthusiastic screening results in
diagnosing diabetes in people that do not have it. Since non-diabetic
individuals have a more favourable life course than persons with
diabetes, this difference in the outcome may be erroneously attributed
Randomized controlled trials (RCTs) are usually regarded as the best
means to evaluate the effectiveness of screening and early treatment.
They are superior to observational studies because, if randomisation is
successful, the possible confounding effects of individual attributes and
health-related behaviours other than the decision to take up screening
can be eliminated89 .
One of the reasons why data from RCTs that apply available treatment
to a screened group, but not a control group are unavailable, is that
such studies require long-term follow-up of a large number of
participants. The feasibility of RCTs is further decreased by the need to
account for people who refuse to participate, as well as for people in
the control group who are offered and accept screening outside the
programme. Nevertheless, RCTs which can take into account these
issues are potentially feasible and should be encouraged.
8.2. The need for observational studies
RCTs to evaluate screening for diabetes have not been conducted so
far, and even observational studies are scarce, in contrast to screening
several other chronic conditions, particularly some of the cancers.
Once a screening method has been shown to be effective in an RCT,
cohort studies in the general population could measure how a
particular screening programme performs in a specific population. A
cohort study could demonstrate the rate of particular outcomes among
participants, refusers and non-invited subjects or screened versus non-
In contrast to RCTs and cohort studies, case-control studies estimate
the individual's risk reduction if screened, although the protective
effect of the screening procedure itself is not quantifiable. Although
case-control studies are often regarded with scepticism on account of
length-time, lead-time and selection biases, these can largely be
accounted for with appropriate design and analysis 90 . Case-control
studies cannot replace RCTs for the evaluation of the effect of
screening, but they can be used to monitor a programme's effectiveness
once screening has been widely introduced91 .
Recently, the STARD initiative92,93 has made recommendations for the
reporting of studies of diagnostic accuracy. These recommendations
include a checklist and flow diagram for the elements (such as the
numbers of eligible patients, exclusions, abnormal, normal and
inconclusive results etc.) vital to the interpretation of results. By
analogy, standardisation of the reporting of observational studies of
screening would facilitate their interpretation and enable their
generalisability to be more easily assessed.
As a result of discussions at the WHO/IDF meeting, it was decided to
establish the ‘DETECT-2’ collaboration. This will compare and
evaluate selected strategies for screening for undiagnosed type 2
diabetes across a range of populations from diverse ethnic backgrounds
and to establish simple screening strategy options. DETECT-2, by
using population based data from various populations will examine the
prognostic implications with regard to morbidity and mortality for
individuals categorised on the basis of a screening programme for type
8.3. The need for economic evidence
There are only two cost-utility evaluations for type 2 diabetes, one
comparing opportunistic screening and the other comparing mass
screening with routine care94,95. The US study did not include the
benefit of preventing macrovascular complications. The study in
Taiwan demonstrated that screening may be cost-effective in countries
with high diabetes prevalence. There are no studies that have looked at
cost-utility of screening for diabetes in populations deemed to be at
high risk of type 2 diabetes.
8.4. The use of modelling studies
All issues around the effectiveness of screening for diabetes cannot
realistically be resolved by RCTs. The development of less costly and
less time-consuming methods is to be advocated. One study used a
Markov model to evaluate the efficacy of population screening for
diabetes96 , and another used the Monte Carlo model to estimate the
lifetime costs and benefits of opportunistic screening for diabetes
(CDC 1998). Both indicated there could be economic grounds for
screening. Although modelling studies do not provide answers, they
direct attention to the right questions which can then be addressed in
8.5 The need for evidence on the psycho-social effects of early
Screening may lead to over-diagnosis, inappropriate investigation and
treatment, avoidable adverse effects and unnecessary psychosocial and
economic costs97 . However, there are no studies specifically examining
these issues in diabetes. Physical harm associated with screening for
diabetes may be considered negligible, but psychological and social
harm could be more substantial.
A diagnosis of Type 2 diabetes has potential implications for
employment and personal insurance. Treatment with insulin precludes
some forms of employment and insurance premiums are higher for
persons with diabetes. Anxiety caused by false positive results in
screening for diabetes is unlikely to be as high as that caused by false
positive results in screening for cancer. However, since screening
programmes involve a large number of people, even a small adverse
effect on quality of life or health-related behaviour could affect public
health. Therefore, studies of the psychosocial effects of screening in
diabetes are needed to complement studies of effectiveness.
9 Implementing policies about screening for type 2 diabetes
Two examples of national policies, from Brazil and Mexico, favouring
screening for type 2 diabetes are given on the following pages. While not
necessarily advocating these approaches, the examples are provided here as
illustrations of programmes which have been implemented on a large scale.
The Case of Brazil
Brazil recently performed nationwide community screening for diabetes and hypertension as part
of its National Re-organization Plan for the Care of Diabetes and Hypertension. In addition to
detection of undiagnosed diabetes, the purpose of the screening programme was two-fold:
- to raise public awareness of the importance of diabetes and hypertension, and
- to focus the efforts of primary care and health administrators on the restructuring and
capacity building necessary for adequate diagnosis, basic treatment, and prevention
of complications of diabetes within the primary care sector as well as for the creation
of adequate referral networks.
Following an initial countrywide training of over 13,000 health professionals in the diagnosis and
treatment of diabetes, a mass media campaign invited members of the public to participate in
capillary glucose testing in March and April of 2001. Over 5300 municipalities participated in the
During the campaign, 21.8 million (73% of those targetted, adults > 40 years of age) were tested
with glucose meters. Of these, 1% (about 0.25 million) had values > 15mmol/l (270mg/dl) and
were referred directly for medical management. An additional 3% (about 0.61 million) had
glucose screening values above the diabetes cut-off points, and immediately received a referral for
confirmatory diagnostic testing.
An additional 12% (3.4 million) were test positive at lower values (fasting > 5.5mmol/l
(100mg/dl) or non-fasting > 7.8mmol/L (140mg/dl) and were counselled to return within 3 months
for further evaluation. Within 6 months of the screening programme, an additional 1.2 million
fasting glucose determinations were performed by outpatient laboratories of the National Health
System, presumably, in great part, as a result of the diagnostic demand induced by the program.
Evaluation of the process and costs of this programme are currently contributing to Brazil’s effort
to shift diabetes prevention and management out of hospitals and into primary care.
For further details contact: Professor Maria Ines Schmidt (firstname.lastname@example.org
The Case of Mexico: “You Have Diabetes but You Don’t Know it”
In Mexico the 2000 National Health Survey demonstrated a prevalence of diabetes of 10.9%
among those aged 20 years and over. This meant that in 2000 an estimated 5 million people were
suffering diabetes in Mexico.
The Mexican Ministry of Health is conducting continuous diabetes (and hypertension) screening
among those aged 20 years and over contacting their medical services for any reason. Volunteers
are also evaluated during fairs and diabetes prevention activities such as those commemorating
World Diabetes Day in some states. The target population of the government health care plan
includes 41% of the Mexican population (about 41 million people in 2001).
The aim of the screening system is to identify those with undiagnosed diabetes to provide early
treatment and prevent or delay the onset of long-term complications. It also focuses on the
identification of those at high risk of presenting diabetes, aiming to decrease the frequency of
known risk factors such as obesity, lack of physical activity and deficient diet. The screening
process is divided in two phases. The first one is the identification of individuals at high risk of
diabetes through the application of a questionnaire named “You have diabetes but you don’t
know it”. This questionnaire has seven questions and includes the calculation of BMI and the
measurement of the waist circumference. The questionnaire was validated previously for the
Mexican population. During the second phase, those obtaining scores of 10 or more points in the
questionnaire are tested for blood glucose. In cases with capillary fasting blood glucose of 100
mg dl or capillary non-fasting blood glucose of 140 mg dl, a confirmatory test is required.
Confirmation of the diabetes diagnosis includes an Oral Glucose Tolerance Test (OGTT) if
Those with newly diagnosed diabetes and also those at risk are referred to different services
included in the health system to commence diabetes education and treatment (if indicated). They
are also invited to participate in a social group of people called “Mutual-help Group”. Those who
obtained scores lower than 10 points in the questionnaire receive counselling about maintaining
adequate weight, diet and physical activity. The first evaluation of this plan in 2000 included a
pilot of 6,186 persons in four states, 43% were considered at high risk of presenting diabetes and
1.6% was diagnosed with diabetes.
The cost of the screening was estimated at US$8.36 per newly diagnosed person with diabetes.
In 2001 overall 3,945,885 people were evaluated with the application of the questionnaire,
572,153 people were tested for blood glucose and a total of 273,149 people from 32 states were
identified as newly diagnosed with diabetes. Results for 2002 showed that a total of 3,985,860
were evaluated through the application of the questionnaire, 576,825 blood glucose tests were
conducted and 313,124 people were diagnosed with diabetes.
For further details contact the Health Program for the Adult and the Elderly of Mexico, at the
National Center for Epidemiological Surveillance (email@example.com or
10 Conclusions and recommendations
1 The issue of screening for type 2 diabetes is important both in
terms of individual health, day-to-day clinical practice and
public health policy.
2 There is currently no direct evidence* as to whether individuals
will or will not benefit from the early detection of type 2
diabetes through screening.
3 Despite this lack of direct evidence, early detection through
screening is already taking place both by inviting individuals
from the general population to come forward for screening and,
opportunistically, when individuals perceived to be at high risk
of developing diabetes attend for health care (usually primary
health care) for other reasons.
4 These activities present opportunities for collecting
observational data which, although no substitute for direct RCT
evidence, can provide important, circumstantial evidence about
efficiency, costs and impact.
5 There is direct evidence that the incidence of diabetes can be
reduced in people at high risk of the future development of type
2 diabetes who may be identified as a result of activities
directed towards diabetes detection.
6 If screening can be shown to be beneficial, the most important
epidemiological considerations determining whether to screen
in any given population will be (1) the prevalence of
undiagnosed type 2 diabetes in that population and (2) the
degree to which type 2 diabetes is associated with risk of
cardiovascular disease, diabetes specific complications and
other important health outcomes in that population.
7 The most important health systems considerations will be its
capacity (1) to carry out the screening (2) to provide effective
health care for those who screen positive (3) to address the
psycho-social needs of those who undergo screening and (4) to
implement effective prevention in those who, though not
confirmed to have diabetes at the time, are at high risk of its
8 The most important population considerations will be (1) the
acceptability of the screening programme to those invited to
‘Direct evidence’ is that from randomised controlled trials (RCTs) specifically
designed to answer questions related to early detection through screening.
attend (2) the extent to which any lack of acceptability reduces
uptake (3) the psychosocial impact of each screening outcome –
positive and negative, ‘true’ and ‘false’ and (4) the ability of
those found to be at risk of future development of diabetes to
modify these risk.
9 The most important economic considerations are (1) the cost of
early detection to the health system and to the individual (2) the
extra costs of treatment following early detection and (3) the
relative cost effectiveness of early detection compared with that
of improving the care of clinically detected (as opposed to
screen detected) cases.
10 The most appropriate protocol for screening for undiagnosed
type 2 diabetes in a particular setting should consider (1) the
sensitivity and specificity of the screening methods available
(2) the number of people who will need to be screened (3) the
number of people who will need subsequent diagnostic testing
(4) resource implications and (5) costs.
11 Screening for type 2 diabetes is a dynamic topic in which new
evidence will become available and further considerations will
arise over time.
10.2 Recomme ndations
1 Health authorities and professional organisations should
formulate policies concerning screening for type 2 diabetes
even if the policy is that screening is not currently to be
advocated. In formulating that policy, the benefits and costs to
the individual and their well-being are of paramount
2 There is an urgent need for direct RCT evidence on the effects
of early detection of type 2 diabetes through screening. Such
evidence should include health outcomes related to diabetes,
cardiovascular disease, psychosocial outcomes and economic
considerations for individuals, health systems and the wider
society. Although RCTs directed to answering these questions
may be costly and logistically difficult, there is, in the current
state of knowledge, no ethical reason why they should not be
3 Since the results of such RCTs will not be available for some
time (if ever), there is also an urgent need to develop a
framework (or model) which would permit countries to
evaluate the cost-effectiveness of earlier detection of diabetes
compared to other preventive and therapeutic interventions.
4 Testing apparently unaffected individuals at increased risk of
having diabetes when these individuals attend for health care
for other reasons (sometimes called ‘opportunistic screening’)
may be justified provided (1) the reasons for testing are
adequately explained to the individual (2) the health system has
the capacity for the clinical management of those who screen
positive (3) methods with adequate sensitivity and specificity
are available (4) the psycho-social needs of those who screen
positive and those who screen negative can be met and (5) the
health system can implement effective preventive strategies for
those confirmed to be at high risk for the development of
diabetes. There is no evidence to justify haphazard screening.
5 If such opportunistic screening is advocated then this should be
carried out according to a policy which should (1) be clear and
relevant in its aims and objectives (2) be based as far as
possible on sound evidence (3) take into account the
epidemiology of type 2 diabetes and related cardiovascular
disease risk in the population and (4) be sensitive to competing
local health priorities.
6 The choice of the method or methods for screening will depend
on the resources available, the acceptability of the methods in
the population being screened and the levels of sensitivity,
specificity etc. that are required. Methods of screening which
might be regarded as unacceptable in high resource settings
(e.g. testing for urinary glucose) may be suitable in low
7 Where screening is already taking place, formal evaluation
should be integral to these activities. The results of such
evaluations could contribute to the general assessment of the
value of early detection and should be used in the modification
or curtailment of the activities being evaluated.
8 Given the dynamic nature of this topic, policies for screening
for type 2 diabetes must be reviewed from time to time as new
Annex 1 Participants of the WHO/IDF meeting (9-11 May 2002)
Professor Sir K.G.M.M. Alberti, Professor of Medicine, Department of Diabetes and Metabolism,
The Medical School, Newcastle-Upon-Tyne NE2 4HH, United Kingdom.
Professor S. Colagiuri (Chair), Director, Diabetes Services, Diabetes Centre, Prince of Wales
Hospital, Randwick NSW 2031, Australia.
Dr E. Goyder, Public Health Section, SCHARR, University of Sheffield, Regent Court, 30 Regent
Street, Sheffield, S1 4DA, United Kingdom.
Dr W. Herman, Division of Endocrinology and Metabolism, Department of Internal Medicine, 3920
Taubman Center, Box 0354, Ann Arbor, MI 48109, USA.
Professor D Johnston, Professor of Clinical Endocrinology, Imperial College School of Medicine, St
Mary’s Hospital, Praed Street, London W2 1NY, United Kingdom.
Dr N. Levitt, Endocrine/Diabetes Unit, department of Medicine, UCT Medical School, Old Groote
Schuur Hospital, J-Floor, Obxservatory 7925, Sth Africa.
Ms J. Lindstrom, MSc Researcher, Diabetes and Genetic Epidemiology Unit, Department of
Epidemiology and Health Promotion, National Public Health Institute, Mannerheimintie 166, FIN-
00300 Helsinki, Finland.
Dr A. Ramachandran, Diabetes Research Centre and MV Hospital for Diabetes, 4 Main Road,
Royapuram, Chennai – 13, India.
Dr M. I. Scmidt, Faculdade de Medicina, Universidade Federal do Rio Grande do Sul, Ramiro
Barcelos, 2600/414, Porto Alegre-RS-Brazil, CEP 90210.
Dr K. Siddiqi , Nuffield Institude for Health, University of Leeds, 71-75 Clarendon Road, Leeds, LS2
9PL, United Kingdom.
Dr N. Tajima, 3rd Dept. of Internal Medicine, Jikei University School of Medicine, 3-25-8
Nishishinbashi, Minato-ku, Tokyo 105, Japan.
Dr N. Wareham, Department of Community Medicine, University of Cambridge, University Forvie
Site, Robinson Way, Cambridge CB2 2SR, Royaume Uni, United Kingdom.
Professor Zhu Xi-Xing, Hua Shan Hospital, 12 Wulumuchi Zhong Road, Shanghai 200040, People’s
Republic of China.
Dr A. Barcelo
Dr R. Bengoa
Dr H. King
Dr J. Leowski
Dr G. Roglic (Co-organiser)
Dr R. Williams (Co-organiser and Facilitator)
Dr D. Yach
Annex 2 Acknowledgements
The co-organisers of the WHO/IDF meeting would like to acknowledge the
contributions made by a number of individuals, both those who attended the meeting
and those who have commented on various drafts of this report. Particular thanks are
due to Professor Colagiuri who chaired the meeting and who has made extensive
contributions to the text of the report. We would also like to acknowledge the
assistance of Marie-Christine Nedelec in organising the meeting and Sally Belcher in
the preparation of the final manuscript.
Table 1 – Biochemical criteria (venous plasma) for the diagnosis of diabetes, impaired glucose tolerance and impaired fasting glucose or
impaired fasting glucose*
Glucose concentration, mmol l-1 (mg dl-1)
Fasting and/or => 7.0 (=> 126)
2-h post glucose load => 11.1 (=> 200)
Impaired glucose tolerance (IGT)
Fasting (if measured) < 7.0 (< 126)
and 2-h post glucose load =>7.8 (=>140 ) and < 11.1 (< 200)
Impaired fasting glycaemia or impaired fasting glucose (IFG)
=> 6.1 (=> 110) and <7.0 (<126)
and (if measured)
2-h post glucose load < 7.8 (< 140)
*adapted from "Definition,Diagnosis and Classification of Diabetes Mellitus and its Complications, WHO Geneva, and the “International Diabetes Federation IGT/IFG
Consensus Statement”1,2 . Venous plasma, venous whole blood and venous capillary values are given in the original reports.